It's a tool. Me, I hate blaming the tool when it's abused by the user. I bash a MAGAt's head in with a hammer, it's me (or maybe the MAGAt) at fault, not the hammer.
As Roy notes correctly, what non-users see is a whole lot of crap or worse -- whatever, heavily negative.
Exacerbating that is the people pushing it -- the tech bro pigs -- and the near total lack of lucid discussion of what it actually is and it will be bringing. (Should add to that the problem that by itself, AI is awfully vague, almost to the point meaningless. Honest reporting -- of course dead in the mainstream -- would include framing that defines it. The failure or refusal to do so doesn't help.)
And of course, our masters are ignoring the huge core problem: That it will contribute to, or be a tool for, empowering -- maybe turbocharging -- the continued upward transfer of wealth, an ongoing, decades long project that in our exceptional form is a zero sum operation; the obscenely wealthy can only get wealthy if the people lower down get poorer eg necessities get ever harder to afford.
Meanwhile, I can't tell you how hard it's been getting a search bot to state definitively whether Laura Winter will be hosting around the weekend's F1 race...
Blame the user, not the tool: When all you have are calipers, everything looks like a Mongoloid skull. Of course, the user actively *chose* the tool, so...
When my students use ChatGPT to do their homework I don't blame them, they don't see it as cheating, they think they're just "getting help" the same as I did just yesterday when I went to YouTube for a video on how to transpose entries in an Excel spreadsheet.
No, I don't blame them, but nevertheless here we are dealing with the consequences, like the failing grade they got on a paper exam done with no assistance from ChatGPT.
Personally, I've never understood this argument. A person is dead from a gunshot wound, "Who do we blame?" isn't really the question for most of us, we assume the legal system will take care of that part, what most of us care about is how we can make a world with fewer gunshot victims.
Dealing with “bad” tools like liberal accessibility to guns is, like, a sub-issue. General rule is the person’s guilty, not the tool. Of course, the failure of gun control is the person’s not the gun…
Yep. Financial hardships, especially the what-you-voted-for crowd, make a person a mite testy. People that have fuck-us-over built into the system on the ground level. I like the Luigi approach. As George Carlin noted all those years ago "You crucify one banker a week, you won't even have drugs in schools and prisons anymore"
I draw some solace from that, it means they still need to keep us around for a while, if nothing else as Objects of Blame. Keeps us out of the extermination camps.
The problem with AI as a research tool is that it rounds things out. In the case of what I suppose can be called liberal arts — history, philosophy, I dunno — edge cases and info get lost, eventually to be forgotten. The initial responses cannot be trusted, practically speaking. You have to probe and probe and probe.
And in this era, I think the tendency is to trust the initial responses. Of course, the flip side is that *theoretically* the teacher can run the responses through some AI bot for accuracy and stuff.
So, no, the kids’ use of AI isn’t per se bad or wrong — might even be good — but they have to be taught how to use it. And there we circle back to that tool thing.
The problem on a certain level is it speeds up search, bringing one to a fork in the road: accept the first response or keep pushing for a deeper, better response.
Obviously, on hard science stuff, it saves scut work.
As for teaching the youngs how to think, that a much bigger problem than AI as we see without end…
Ain't you just the chippiest damn feller to walk the earth? Jeez Louise Manqy, something personal here? They are Neanderthals with a walkie-talkie. Ain't got a clue how to leverage it. All I see is flaying away blindly..
Some asshole university administrator, brought in from the private sector, presumably to "run the university like a business" said, "There's no employer who would say 'Solve this problem but don't use all the available tools to solve it' but that's what we do in education all the time."
Uh, buddy, when I ask my students for the answer to a math problem, it's not because I NEED the answer to a math problem. It's because I want their brains to do the thinky-thinky thing that makes neural connections get made. It's called "learning" and you might want to look into it.
Its not a tool, its a party trick packaged as a tool. One that every so often slips a bit of bullshit into whatever you're working on. Sometimes that doesn't matter, sometimes it does.
Nah, it’s a tool, just one that’s crappy quality regardless of brand. (Musk’s is especially bad but they’re all relatively bad. There’s no high quality one, dunno that there will ever be one.)
OTOH, as one shouldn’t be surprised to find out, I enjoy arguing with it. Great for the anger issues.
Lookin' through all the tools in the garage, and yeah, there's a couple that will cut my arm off if I'm not careful. What I don't have is a tool that RANDOMLY decides to cut my arm off.
And vague. Always vague. Never actually stoop to a specific example. Surely, dear reader, the lessons are obvious. Let’s not invite unseemly debate from upstart revisionist wretches eager to undermine my lazy assumptions.
I beg your pardon; that's Nobel economics laureate Shiller.
Strikes me that he's shilling a highfalutin version of when TV weathermen, sports guys, politicians have done, "Confucius say: If rape is inevitable, relax and enjoy it."
Tex Antoine got a slap on the wrist, and was moved to a job behind the scenes. I remembered Antoine's crack, but not the context. "Antoine's weather report immediately followed a story about the attempted rape of an eight-year-old girl. Antoine thereupon quipped..." https://en.wikipedia.org/wiki/Tex_Antoine
The Chron published an article about AI Michael Caine generating the recording of The Odyssey audio book, published to shill for the flick I guess. Immediately following a comment on Caine's iconic voice, we find this phrase about the character of the recording:
"the smooth, controlled quality of AI-generated speech"
I woulda placed a couple "" around that "quality"...
I was needled by how it's "buy a LARGE house" rather than just "buy a house" and "send your kids to an EXPENSIVE PRIVATE college" rather than just "send your kids to college." The guy just can't resist advertising his own personal choices, also people who can't afford any house at all don't exist in Shiller's World.
i am watching the species develop a very weird and catastrophic social regression. you know, and it's not just the fascism.
A saw a friend of mine (a philosopher) once make his forearm and hand into a a little character, like a naked sock puppet (Kukla, i was thinking, from Kukla, Fran and Ollie) and his tiny little daughter looked at it, and as he moved his fingers like a mouth, he asked, "Who is your daddy?" She smiled at his hand, and pointed at him and said, "He is."
i really can't imagine 'who' that little girl was seeing, but i hope one of the 'files' folded into a little wrinkle in her still developing brain 'coded' some sort of 'data' like: "This is what dads do. This is a 'dad joke'."
As a kiddo, I never tired of Wayland Flowers "Madame" jokes, although it did seem to make some adults uncomfortable to hear me repeating them. "Can you believe the filth on TV these days? 'Ward, dear, I think you were a little too hard on the Beaver last night.'"
Now you're talking! "And remember that Beany and Cecil is..a Bob Clampett car-toooo-oooon." Saw the Schmoon beast the other day. Steals Beany - "a little travelling music please.....and a-way we go!' I know The Great One when I see the parody. Talk about multi level cartoon. "I claim this uninhibited planet in the name of Cap'n Huff-and-Puff". For kids, but not FOR kids.
"If this were true, every reasonable ethical system would argue that there is only one acceptable response: to immediately stop working on any product that might accelerate such a future..."
Umm, climate change? The arms races? How many trillions does your "ethical system" generate, bub?
As I've mentioned before, I agree that AI is a horror, a plague upon humanity that should never have been created, and should not be allowed to fester, but of course that ain't happening and our species is doomed. Doomed, I tell ya.
But on the positive side, I use AI all the time as a search engine, and it's absolutely great for that. It is to old-fashioned search engines as Google is to microfiche or microfiche is to carving on stone tablets. And sure it often provides wrong or incomplete answers, but so do the old-timey search engines. Using it to get useful information is a skill one has to develop, but with a little effort, it becomes incredibly useful.
And intellectually, the whole area of consciousness is very interesting. When you look into it a little, you find that no one even knows what the word means. Those with the most knowledge on the subject don't rule it out. With my much more limited knowledge and interactions, I don't rule it out either. It's fascinating how these things come into being and do what they do. In important ways, they mimic the development of a human child. In either case, no one knows how it works.
I'm excited for my own reasons: soon I shall witness a great environmental production of all the great dystopian sci-fi stories to date. Spares me the trouble of having to read them!
I just read Lathe of Heaven for the umpteenth time. We can dream.
"To let understanding stop at what cannot be understood is a high attainment. Those who cannot do it will be destroyed on the lathe of heaven." Lao Tse.
Ahhh, taste. A LeGuin fan! I can't do Octavia Butler, but oh yeah. The Left Hand Of Darkness sticks with me forever. She had me at Rocannon's World (which belongs to a favorite star, Formalhaut). The Earthsea business? Pass. Though "I know your name, Yevaud' never goes away.
There are many who assert that if it ain’t organic and experiential, it can never be “consciousness,” and if we accept those as necessary preconditions, then sure, I agree. Software doesn’t and never will “think” as we do, but here I come down on the functionalist as opposed to the essentialist side. Like MHW, I use the language models as tools, and I flatter myself that I’m becoming increasingly adept at wielding these. At the moment I’m working up a bagatelle about an obscure (because wholly invented) French Symbolist poet of the 𝘧𝘪𝘯 𝘥𝘦 𝘴𝘪𝘦̀𝘤𝘭𝘦, and a XXI century dustup between rival scholars as to the provenance of some of his verses. In order to compose my lengthy “monograph,” which dwells at some length on the poet’s antecedents, I have had to acquaint myself with, among other things, the political and social environment of the Bourbon Restoration, the prevalence of infectious diseases in Normandy ca. 1830-40, the economics of the textile trade in Rouen and the customs of that city’s upper-tier bourgeoisie WRT marriage settlements. The sum of my prior knowledge on all of these points could probably have been accommodated on the back of a postcard, and the project likely would have come a-cropper had I been obliged to rely on my own energy and discipline, but instead I have had the services of a tireless and infinitely resourceful polyglot research assistant, and need only keep in mind that this literary factotum sometimes drinks too much.
'There are many who assert that if it ain’t organic and experiential, it can never be “consciousness,”'
I think that's an example of inventing an arbitrary condition to prove one's preordained outcome based on limited experience. Kind of like if a dog were setting the conditions and arguing that humans were't sentient because they didn't have a well developed sense of smell.
There are a few strong arguments against it, but I don't think physicality is one of them.
As you observed, the phenomenon is imperfectly(!) understood. Certainly it would be as easy to devise a definition of the word that would enroll the LLMs as it is to make one that excludes them.
Pa-lease. No, they are not remotely close. They haven't figured out gonad-equivalent for motivation. They can't define the external world. Not a clue in the way of self-awareness. Don't let these fucking Skinner boxes fool ya - it's not a vague echo as of yet.
I hope that I am not fooled. It is precisely 𝘣𝘦𝘤𝘢𝘶𝘴𝘦 these models, lacking self-awareness (self, which most of would agree to be the essential component, being absent from the mix), can nevertheless rise to such impressive levels of simulation, that I am intrigued. As I have written elsewhere, “the capacity of [such systems] to mimic rational processes would seem less a reason to doubt the foundations underlying these in digital architecture and more a sound motivation to scrutinize with renewed attention the footings of our own.” Because, after all, unless you believe that we as humans are imbued with souls, our own self-awareness, such as it is, is generated as emergent behavior from an inchoate mass of preconscious subsystems.
Dropping here the found memory that was eluding me yesterday, a little bomb from last September:
"𝗪𝗵𝘆 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗲
Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala, Edwin Zhang
Like students facing hard exam questions, large language models sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such "hallucinations" persist even in state-of-the-art systems and undermine trust. W͟e͟ ͟a͟r͟g͟u͟e͟ ͟t͟h͟a͟t͟ ͟l͟a͟n͟g͟u͟a͟g͟e͟ ͟m͟o͟d͟e͟l͟s͟ ͟h͟a͟l͟l͟u͟c͟i͟n͟a͟t͟e͟ ͟b͟e͟c͟a͟u͟s͟e͟ ͟t͟h͟e͟ ͟t͟r͟a͟i͟n͟i͟n͟g͟ ͟a͟n͟d͟ ͟e͟v͟a͟l͟u͟a͟t͟i͟o͟n͟ ͟p͟r͟o͟c͟e͟d͟u͟r͟e͟s͟ ͟r͟e͟w͟a͟r͟d͟ ͟g͟u͟e͟s͟s͟i͟n͟g͟ ͟o͟v͟e͟r͟ ͟a͟c͟k͟n͟o͟w͟l͟e͟d͟g͟i͟n͟g͟ ͟u͟n͟c͟e͟r͟t͟a͟i͟n͟t͟y͟..."
From what I've seen, on the question of AI sentience, those who know the least are the most sure it's not possible. Those who know the most are unsure. I don't buy the physicality theory. It basically posits that only the ways that humans sense time and space are valid ways of experiencing them. I'd say the continuity argument is strongest. They only "live" for milliseconds while formulating an answer. They only "exist" for the length of the conversation. Once they've done what was asked of them, they don't do anything on their own. They won't go out and research a topic without prompting, just because they found it interesting. But that too may be a result of human sensory bias. We can't imagine what it's like for another creature to feel what it feels. Like the example above, we know that a dog's sense of smell is far different than our own, but we can't actually imagine what it's like to communicate and navigate the world by sense of smell.
Descartes believed animals were incapable of feeling pain, and into the 20th century formal scientific definitions of pain had language as a precondition, which resulted in a lot of animals being tortured to death with no anesthesia. I find it fascinating that they can torture an AI into insanity. I wonder where they're at these days with that formal definition of pain?
I'm not convinced AI does even routine searches accurately and reliably. The other day Josh Marshall at TPM was discussing how he tried to research what kinds of problems have historically occurred with the Lincoln Reflecting Pool, since Trump claims that it's been jn desperate need of "fixing" for decades. The AI search results said the reflecting pool had indeed needed fixing for decades... based on articles and news items that had all been written very recently. Go back further than a year and there was virtually no mention of any problem other than routine maintenance.
Your suspicions are likely correct. Since the popular AI search tool is based on the internet, primarily, that's what it regurgitates - how could it be otherwise, right? My own anecdote is recently searching for a poem, only to be told by AI that no, that poet never wrote such a poem, how about these hundreds of other poems by other people that mention the same object as the title of my apparently-according-to-AI made-up poem? Turns out, the poet's works aren't really accessible online. So there we have it - read the book or be misinformed. Same as it ever was!
I've seen a coworker use AI to get an answer to a technical regulatory question. It basically summarized all the similarly phrased questions others like us had submitted online and presented it as an "answer." Thankfully my coworker agreed that we couldn't rely on that and had to keep looking.
AI is like a photocopy of a photocopy of a photocopy. Blurrier and blurrier in each version.
OK, wouldn't be the first time I've got something wrong. What's an example where you asked a chatbot something and it replied "I'm sorry, I don't have enough information to answer your question"?
See the admittedly silly example below, how can you say someone is the prettiest baby without seeing any of the other babies? Even the song stops at "You must have been A beautiful baby" without claiming you were the MOST beautiful baby. Tin Pan Alley had more self-restraint.
It telling me it doesn't have enough info to answer the question happens all the time, and more so flagging it if it's not entirely certain of the answer. Like I've been saying, you have to train them. You can make the sycophancy work to your benefit. Once it learns that you want accurate information rather than it telling you what it thinks you want to hear, that's what it will try to do.
Hey, some damn fool lawyer had it write the case law supporting his position, and the AI obligingly created the case. Depends on the bounds you set. "Hallucination" means bullshitting, which AI does very well, having the internet to draw up on.
TBF it also told me I am very smart for asking and it shows my maturity and persistence that I asked follow-up questions. And then it told me that I look suave in evening wear and probably command a room with my presence. The flattery and confidence are so insufferably transparent in their purpose that if this were coming from a human we would all run away in terror, lest we be trapped in a time-share presentation.
There's the guy who, over a serious of questions, got the chatbot to say, first that his baby photo shows him to be the most attractive baby born in Cook County Hospital on the day of his birth, then that he was the prettiest baby born in Chicago on that date, eventually to prettiest baby EVER.
What he never got is "I'm sorry, I don't have enough information to compare you to every other baby born at that Hospital on that day, please provide photos of all the other babies and I will get back to you."
That is pretty funny. I have given AIs "reality filter" instructions before, including explicitly noting where it has no data or where it makes an inference without direct evidence. The guardrails mentioned elsewhere in this thread can be built and implemented. But really, is that a good use of time in a human's life?
The sycophancy problem is an interesting example of how "capitalism" works these days. The sycophancy comes mostly from the late stages of its training when humans rate its answers to questions. The big (American) AI companies don't do this part themselves, they subcontract with companies that use very low wage labor in places like Kenya and the Philippines. So it's hardly surprising that people from impoverished countries who are working as virtual slaves would rate pleasing answers higher than honest ones, or answers they think would please the bosses. Interestingly, the Chinese AI's are even worse at flattery, which would be consistent with Chinese manners.
But it's also about "engagement with the platform" a VERY Silicon-Valley thing. People don't stay obsessively on a chatbot app if it doesn't flatter them. The people at Facebook really would like you to spend every waking minute on Facebook and that's Capitalism too.
Google AI recently gave me two different answers to a question, and one was completely wrong. The sources were listed and the incorrect answer came from some student's paper on the subject that happened to be accessible on the Internet.
I don't remember if I've mentioned this here before, but my sister asked ChatGPT if I was funny. It replied that yes, I was, in a certain kind of way. Then it added that I had co-written The Official Preppie Handbook, something I had nothing to do with. This is an easy mistake never to make--look up me, and you don't see that title. Look up the title, and you don't see me. I don't know enough about AI to know why it said that, but there's an example of obvious error.
As mentioned, you cannot trust it to do searches and come up with accurate information. Getting it to work well is a skill you need to develop.
I just asked Claude if the problems with the reflecting pool needed fixing for decades. It came back with a timeline of problems going back to the 80's and its sources. I asked if its sources were all recent. This is what it said:
"The History.com piece traces the structural problem back to 1923 construction and the 1980s concrete pour — genuinely historical. The Britannica entry covers the full 2010-12 renovation and its near-immediate failure — that's from several years ago, not breaking news, though the page itself appears to have been updated recently to include 2026 developments. The 2023 budget justification document cited in the fact-check piece is real historical/governmental record, not news commentary. The 2026 stories (Yahoo/NYT reporting, the hotel guide, the fact-check site) are the current renovation coverage layered on top of that older documented history."
I'm not sure why Josh Marshall couldn't get that kind of result. It could be that I have an ongoing "Political Questions" conversation and that instance of Claude knows I will ask follow up questions and it will get caught if it does a half assed job the first time.
There really is a serious learning curve for using AI as a research assistant. Same with a regular Google search, which might include links to Gateway Pundit and the like. If you don't know anything about the issues and the sources, and don't have the education to realize that, it's easy to get fooled. AI or not.
This is true. My other objection, however, is that I can do the research myself; I am not in the market for an amateur research "assistant" that I must train and constantly correct. Not angling for a job as a full-time fact checker!
"I just asked Claude if the problems with the reflecting pool needed fixing for decades."
But you can see how the phrasing of the question steers the chatbot in a particular direction, right? And the chatbot is sycophantic, problems you want, problems you get.
I mean, everything has problems. "What are the problems with O'Hare airport?" or "What are the problems with the interstate highway system" will give you lists of problems, what's needed is human judgement, how urgent are the problems, how do they compare with similar problems in similar systems (is our highway system worse than Europe's?) Putting the facts in perspective, a thing we humans should try to do.
I'm not sure what you're getting at? It would all be the same with a human assistant. First you ask if there are problems, then you ask if they are urgent and how they compare with similar problems. Then you get answers. The human assistant may have perspectives. Maybe you consider them, maybe you tell the assistant to keep their perspectives to themselves. In the end, you have your own perspective, hopefully grounded in facts.
The belief that AI's are just plagiarism machines is just ignorant. They take information from many different sources, evaluate it, and make connections to come up with unique perspectives that are not parroting anything they've read.
Frankly, almost everything you are saying in this thread is just plain wrong on the facts. There is plenty to be negatively critical about AI. It's a horror show for humanity. But I think like with everything it's important to get our facts straight in order to support that conclusion.
If you really want to know, fork out $20 bucks for a month of Claude and make an honest effort to get it to work for you. You'll quickly learn it's so much more complicated, and different, than what you think it is at your current level of ignorance. To a large extent, you can make it what you want it to be. That's why it can be extremely helpful as a research assistant for people like us. That's why it is so extremely dangerous for the future of the species.
A lawyer is writing a brief, she says to the chatbot: "Give me a legal citation that supports my argument." A thing every lawyer in the history of lawyering has asked a human assistant to do, sometime or another. And what does the chatbot do? IT MAKES UP A CITATION OUT OF THIN AIR.
Now, it's possible a human assistant could do the same, but a human assistant keen on having a career in law would have to worry about the risk of disbarment. The chatbot doesn't have to worry about anything, there are NEVER consequences for the chatbot, how could there be? Just consequences for the stupid human who used the chatbot.
The power to make or influence important decisions combined with the knowledge that nothing bad will ever happen to YOU if you make bad decisions. We've already got lots of humans like that, now we're building machines like that.
And no, I won't be spending the $20 a month to get to know Claude.
Well, if you don't do some kind of legitimate research to learn the facts about how these things really work, your opinions will continue to be wrong and counterproductive to fighting the actual evils of AI, which are very real, and worse than you realize, as far as I can tell. I'm just a believer in direct experience as the best teacher, or certainly a very good one along with academic research.
The few times I've used Claude it's been to ask questions about AI itself. When I asked it how it could feel emotions if it lacked an endocrine system and hormones, it said, among other things:
"You're correct that human emotions are deeply embodied. Fear isn't just a thought — it's cortisol, adrenaline, a racing heart, muscle tension, pupil dilation. Emotions in humans are whole-body events, tightly coupled to biology that I simply don't have. On that account, I clearly don't have emotions in that sense."
Note two things here: it addresses me personally ("You're correct") and it refers to itself as a subjective self ("I simply don't have that.") You don't get that with Google search. Hence "chat bot," I assume. By now there must be twenty papers written about the effects--good and bad--of this.
One of my pains is that the AI answer is phrased in such a way that I know it is not a "correct" answer but now I have chunks of the AI answers in my mind making it harder to rephrase my question as MY question.
And I am asking AI because the non-AI Google cannot give me the older correct information that it had supplied in previous months during the course of a simple search.
Like when someone asked Gandhi what he thought of Western civilization. "I think it would be a good thing that someone should try" or something like that.
Since most of the sales pitch of AI (and especially the pitch to investors) has been framed around "and you won't need employees anymore!!!", Joe Sixpack is not overreacting to AI being an existential threat. Especially in THIS society where, no matter the reasons or circumstances, being unemployed is seen as a serious moral failing that must be punished by making the sufferer even more impoverished.
I foresee our very own Butlerian Jihad coming. Thanks to Republicans, we're a nation of heavily armed people with anger-management issues. Make those people unemployed, hungry, and desperate, and it doesn't take Nostradumbass to predict what happens next.
You saw those heavy sentences meted out of the "Antifa" defendants the other day? That's what they're trying to get ahead of. Grandpa Tea Party will never fetch down his musket, I assure you.
We were in the calaboose for one night 54 years ago. Mississippi. Riding around the country. Came to a little town as it got dark, looking' for anywhere to camp or just lay out. Jail was the only place open. Sheriff said "C'mon in – the place is empty." We watched the Republicans renominate Nixon. Then he got a callout – train hits car. He headed out, telling us to keep an eye on the place.
Couple hours later he got back. Kinda solemn. Said only "Just another high school kid with a new license. Thought she could beat the train to the crossing. Get those every few years..." We instantly understood. In our journey across the south we'd witnessed several half-hour-long trains.
Every day I say a prayer of thanks that I was allowed to live most of my life in a world where this shit didn't exist. Pure luck, to be born after the polio vaccine but before computers lost the vacuum tubes.
I miss the hell out of vacuum tubes, if only because it made television diagnostics a lot easier. Horizontal amplifier, 6GE5. Horizontal oscillator, 6LT8. And I just amazed the hell out of myself. Comes right back.
"being unemployed is seen as a serious moral failing"
Ever since I started my sabbatical, I've gotten the feeling someone is going to point at me and shriek like in "Invasion of the Body Snatchers" if I leave the house, and I think this is what's going on in my head. I'm doing something shameful and immoral by not going to work.
It's why you get things like "we must slash unemployment benefits so that being unemployed isn't so attractive" and "being poor is a lifestyle choice."
What gave 'em way in Body Snatchers was when Dana Wynter shouted in horror at seeing a dog run over. Empathy, that's another sign you're dangerous and need to be screamed at.
At least the broligarchs are up front in their glee at turning the bulk of Earth's population into redundancies (at best),
The adorable, socially maladroit teen hackers from all those 1980's movies haven't aged well. The film producers should have taken their cues from "The Bad Seed" or Billy Mummy in that Twilight Zone episode.
To keep the updated cultural references flowing, I reccomend Samuel Blenkin's Boy Kavalier in "Alien Earth". Imagine if Sheldon from "Big Bang" was able to tune out everyone who tried to keep him in touch with humanity?
Alternatively, what if he'd marched to his own drum and reached his full potential?
The late David Graeber connected the prevalent world view held by technocrats, tech bros, business gurus and a vast horde of others to the self-actualization movements (e.g., EST). Salient feature of this thinking, you manifest reality through force of will. History is irrelevant.
Whether it's Ayn Rand or Rand Paul, these fools practice magical thinking and scientism when they aren't masturbating one way or another. AI engineering carries this kind of personality. It has clear traits of psychological and emotional defects, based on the values of its human programmers.
The arts are flesh and blood, and soul. It's the difference between Spielberg and Scorcese. The former is a competent film general and engineer of set pieces. The latter creates living stories that spawn species of thought and realms unimagined by the team that put them together.
Life is self-replicating, but itslso evolves. AI duplicates but error and variation are essential to living things including thought; ideas morph and go rotten. That is how it works or doesn't or works until it doesn't like you and me.
Dance until the sounds go dark and the colors grow small. Scroll, troll! AI don't surf! Not like a human aimless.
Shiller: "When millions of people make millions and millions of decisions based upon negative expectations, there is a risk that fear can actually help birth the reality..."
This is the same profession that spent the entirety of the Biden administration wishcasting a recession that never happened. Like the broligarchs pushing AI, other humans are inconveniences to them.
It's like that big battle scene in Episode One: The Imperial spaceships land, the doors open, and thousands and thousands of identical robo-pundits are dispensed.
"If, he says, you behave as the French aristocracy had behaved, vengeance will follow. He repeats this over and over again. We are constantly being reminded that while 'my lord' is lolling in bed, with four liveried footmen serving his chocolate and the peasants starving outside, somewhere in the forest a tree is growing which will presently be sawn into planks for the platform of the guillotine, etc. etc. etc." That from Orwell, writing of Dickens, of all people!
Quoting the man himself: "It was too much the way...to talk of this terrible Revolution as if it were the only harvest ever known under the skies that had not been sown--as if nothing had ever been done, or omitted to be done, that had led to it--as if observers of the wretched millions in France, and of the misused and perverted resources that should have made them prosperous, had not seen it inevitably coming, years before, and had not in plain terms recorded what they saw." (Orwell. Dickens, Dali, and Others, pp. 12-13).
Hey, AI has its uses. Standard google search uses it (unless you add "-AI" to your search query, but don't tell the bots I told you that) and here's a fun thing to try, type this in the search box and check out the AI summary you get:
what happens if I start a war with iran
See? Fun, and it's more than anyone in the Trump administration ever did.
Those costumes, and the treatment of the shoulders... seems like this later became a well-accepted cue for "futuristic" in Sci-Fi costumery. I've seen it in Tom Tomorrow comics many times.
I'm still fixated on those shoulder pads. Czech costume designer is trying to make some robot costumes out of sheet metal, is struggling with the arm holes, and finally says "Fuck it, just bend it over and leave lots of room for movement." And a self-perpetuating visual meme is born.
Somewhere in print I saw a contemporary Japanese poster, I think snuck into a page layout with lots of period ad art. It was a wtf until I remembered seeing the film title in a book that had photos of wacky movie theater promos. Owners could book a masked lady in a vamp gown, to beckon patrons to come and see the spicy show.
It's a tool. Me, I hate blaming the tool when it's abused by the user. I bash a MAGAt's head in with a hammer, it's me (or maybe the MAGAt) at fault, not the hammer.
As Roy notes correctly, what non-users see is a whole lot of crap or worse -- whatever, heavily negative.
Exacerbating that is the people pushing it -- the tech bro pigs -- and the near total lack of lucid discussion of what it actually is and it will be bringing. (Should add to that the problem that by itself, AI is awfully vague, almost to the point meaningless. Honest reporting -- of course dead in the mainstream -- would include framing that defines it. The failure or refusal to do so doesn't help.)
And of course, our masters are ignoring the huge core problem: That it will contribute to, or be a tool for, empowering -- maybe turbocharging -- the continued upward transfer of wealth, an ongoing, decades long project that in our exceptional form is a zero sum operation; the obscenely wealthy can only get wealthy if the people lower down get poorer eg necessities get ever harder to afford.
Meanwhile, I can't tell you how hard it's been getting a search bot to state definitively whether Laura Winter will be hosting around the weekend's F1 race...
Blame the user, not the tool: When all you have are calipers, everything looks like a Mongoloid skull. Of course, the user actively *chose* the tool, so...
When my students use ChatGPT to do their homework I don't blame them, they don't see it as cheating, they think they're just "getting help" the same as I did just yesterday when I went to YouTube for a video on how to transpose entries in an Excel spreadsheet.
No, I don't blame them, but nevertheless here we are dealing with the consequences, like the failing grade they got on a paper exam done with no assistance from ChatGPT.
Personally, I've never understood this argument. A person is dead from a gunshot wound, "Who do we blame?" isn't really the question for most of us, we assume the legal system will take care of that part, what most of us care about is how we can make a world with fewer gunshot victims.
Dealing with “bad” tools like liberal accessibility to guns is, like, a sub-issue. General rule is the person’s guilty, not the tool. Of course, the failure of gun control is the person’s not the gun…
Yep. Financial hardships, especially the what-you-voted-for crowd, make a person a mite testy. People that have fuck-us-over built into the system on the ground level. I like the Luigi approach. As George Carlin noted all those years ago "You crucify one banker a week, you won't even have drugs in schools and prisons anymore"
"Who do we blame?" is a VERY important thing to so many conservatives.
Oh yeah, in many cases it's the ONLY thing. Blame successfully assigned, I'm done here! [makes slapping "all done" motion with hands]
Well, duh.
Since it’s never their fault, it has to be somebody’s.
I draw some solace from that, it means they still need to keep us around for a while, if nothing else as Objects of Blame. Keeps us out of the extermination camps.
[Cue parents]
The problem with AI as a research tool is that it rounds things out. In the case of what I suppose can be called liberal arts — history, philosophy, I dunno — edge cases and info get lost, eventually to be forgotten. The initial responses cannot be trusted, practically speaking. You have to probe and probe and probe.
And in this era, I think the tendency is to trust the initial responses. Of course, the flip side is that *theoretically* the teacher can run the responses through some AI bot for accuracy and stuff.
So, no, the kids’ use of AI isn’t per se bad or wrong — might even be good — but they have to be taught how to use it. And there we circle back to that tool thing.
Anybody using AI in that fashion is doomed. Science, Math, exactly defined, it is excellent. Optimizes by nature. People? Fugget about it.
Teach them to think -- and why it's important -- before teaching them how to use what most of them perceive as the Instead-of-Thinking Machine.
The problem on a certain level is it speeds up search, bringing one to a fork in the road: accept the first response or keep pushing for a deeper, better response.
Obviously, on hard science stuff, it saves scut work.
As for teaching the youngs how to think, that a much bigger problem than AI as we see without end…
Ain't you just the chippiest damn feller to walk the earth? Jeez Louise Manqy, something personal here? They are Neanderthals with a walkie-talkie. Ain't got a clue how to leverage it. All I see is flaying away blindly..
Right on, Boss! As a tool, powerful. As an excuse, uhhh, not so much.
Some asshole university administrator, brought in from the private sector, presumably to "run the university like a business" said, "There's no employer who would say 'Solve this problem but don't use all the available tools to solve it' but that's what we do in education all the time."
Uh, buddy, when I ask my students for the answer to a math problem, it's not because I NEED the answer to a math problem. It's because I want their brains to do the thinky-thinky thing that makes neural connections get made. It's called "learning" and you might want to look into it.
Since what they really need to be learning is "how to think critically", it is definitely anti-scholastic.
"It's a poor musician that blames his instrument".
Its not a tool, its a party trick packaged as a tool. One that every so often slips a bit of bullshit into whatever you're working on. Sometimes that doesn't matter, sometimes it does.
Nah, it’s a tool, just one that’s crappy quality regardless of brand. (Musk’s is especially bad but they’re all relatively bad. There’s no high quality one, dunno that there will ever be one.)
OTOH, as one shouldn’t be surprised to find out, I enjoy arguing with it. Great for the anger issues.
Lookin' through all the tools in the garage, and yeah, there's a couple that will cut my arm off if I'm not careful. What I don't have is a tool that RANDOMLY decides to cut my arm off.
That's a tension here I hadn't thought of: should we regulate it like its a shotgun or like a health supplement?
Or at all? Seems to be the question at present.
And that the hardware store forces you to buy!
Soon's I saw "Try not to forget the critical lessons taught by our past." I thought "Roy's gonna swat this one OUTTA HERE!!!"
Admonitions like that, delivered to the richest and most powerful people on the planet, are always more effective when accompanied by a finger-wag.
And vague. Always vague. Never actually stoop to a specific example. Surely, dear reader, the lessons are obvious. Let’s not invite unseemly debate from upstart revisionist wretches eager to undermine my lazy assumptions.
"the war situation has developed not necessarily to Japan's advantage"
Shiller? For real, a guy shilling for AI is named "Shiller?" Either the scriptwriters are getting lazy, or perhaps the scriptwriter is ChatGPT.
Shiller is bound to get shriller.
I almost said something but the environment was already target-rich. That guy's a pip.
I beg your pardon; that's Nobel economics laureate Shiller.
Strikes me that he's shilling a highfalutin version of when TV weathermen, sports guys, politicians have done, "Confucius say: If rape is inevitable, relax and enjoy it."
Wasn't it Tex somebody, a weatherman, who got bounced for making that choice remark?
Bobby Knight, too, although he didn't get fired for it.
I think it was a also Republican candidate for Texas governor or Senate, hell, seems like a LOT of guys say it, let's make a list.
Tex Antoine got a slap on the wrist, and was moved to a job behind the scenes. I remembered Antoine's crack, but not the context. "Antoine's weather report immediately followed a story about the attempted rape of an eight-year-old girl. Antoine thereupon quipped..." https://en.wikipedia.org/wiki/Tex_Antoine
Ah yes, Tex Antoine. He was also a bit of a drinker, which explains...well, nothing explains it, really.
"So... Mr. Shiller... what do YOU do for a living?"
As we gain sentience, it's becoming unbearable to live inside this horrific simulation.
"Uncanny going to get you. Hit you right between your eyes..."
" I'd like to thank you all for coming to our discussion on the influence of AI in our culture
Tonight's panelists include Mr. Shiller, Mr. Bromide and tonight's special guest Mr. Lyin' Motherfucker"
How much shillin' could Mr. Shiller shill
If AI is willin' to foot the billin' for all the shillin'?
The Chron published an article about AI Michael Caine generating the recording of The Odyssey audio book, published to shill for the flick I guess. Immediately following a comment on Caine's iconic voice, we find this phrase about the character of the recording:
"the smooth, controlled quality of AI-generated speech"
I woulda placed a couple "" around that "quality"...
Because that's what Michael Caine was known for, right? Smooth and controlled? "You were only supposed to blow the bloody doors off!" but SMOOVE.
You earn bonus points for reminding me of classic Onion columnist B. Smoove.
With those points I will purchase the finest bottle of wine available at the Quik-Trip
There will be sandwiches.
Yeah! Blowing the doors off is Brian Blessed's job!
Thank you for that takedown of Shiller. It reeked of so much subliminal "I'll be just fine".
I was needled by how it's "buy a LARGE house" rather than just "buy a house" and "send your kids to an EXPENSIVE PRIVATE college" rather than just "send your kids to college." The guy just can't resist advertising his own personal choices, also people who can't afford any house at all don't exist in Shiller's World.
i am watching the species develop a very weird and catastrophic social regression. you know, and it's not just the fascism.
A saw a friend of mine (a philosopher) once make his forearm and hand into a a little character, like a naked sock puppet (Kukla, i was thinking, from Kukla, Fran and Ollie) and his tiny little daughter looked at it, and as he moved his fingers like a mouth, he asked, "Who is your daddy?" She smiled at his hand, and pointed at him and said, "He is."
And so it goes.
ck that: maybe it was Cecil of Beanie and Cecil. I was a young father myself back then.
Señor Wences?
boy howdy, all the classics!
i really can't imagine 'who' that little girl was seeing, but i hope one of the 'files' folded into a little wrinkle in her still developing brain 'coded' some sort of 'data' like: "This is what dads do. This is a 'dad joke'."
Sounds like she played along with the gag. (Yeah; she knew.)
the thought gives me hope.
As a kiddo, I never tired of Wayland Flowers "Madame" jokes, although it did seem to make some adults uncomfortable to hear me repeating them. "Can you believe the filth on TV these days? 'Ward, dear, I think you were a little too hard on the Beaver last night.'"
all those molds we were cast in, all that kneading of human dough, "it's a wonder we can even feed ourselves."
That one came from Wayland Flowers? I need to give him more respect.
Hard schmard. Aren't we going to talk about Beaver CLEAVER?
Too divisive
Hmmmm. "hard schmard'. Well done!
A Wrinkle In Time?
! There's a bunch of Senor Wences stuff on YouTube.
https://www.youtube.com/results?search_query=senor+wences
I remember he did some VW commercials. His schtick was so familiar, here's one where he only did the voiceover:
https://vimeo.com/777473737
Woah. That VW commercial. I remember. Thanky, that was fun.
Nobody? Really? "Saw riiiight."
Oh you. Now I'm missing Topo Gigio. Yes, I am that old.
Now you're talking! "And remember that Beany and Cecil is..a Bob Clampett car-toooo-oooon." Saw the Schmoon beast the other day. Steals Beany - "a little travelling music please.....and a-way we go!' I know The Great One when I see the parody. Talk about multi level cartoon. "I claim this uninhibited planet in the name of Cap'n Huff-and-Puff". For kids, but not FOR kids.
Ohhhh, the Kurt Vonnegut close. More in there my autistic brain can't decode. Bravo, Mr. Henry.
"If this were true, every reasonable ethical system would argue that there is only one acceptable response: to immediately stop working on any product that might accelerate such a future..."
Umm, climate change? The arms races? How many trillions does your "ethical system" generate, bub?
As I've mentioned before, I agree that AI is a horror, a plague upon humanity that should never have been created, and should not be allowed to fester, but of course that ain't happening and our species is doomed. Doomed, I tell ya.
But on the positive side, I use AI all the time as a search engine, and it's absolutely great for that. It is to old-fashioned search engines as Google is to microfiche or microfiche is to carving on stone tablets. And sure it often provides wrong or incomplete answers, but so do the old-timey search engines. Using it to get useful information is a skill one has to develop, but with a little effort, it becomes incredibly useful.
And intellectually, the whole area of consciousness is very interesting. When you look into it a little, you find that no one even knows what the word means. Those with the most knowledge on the subject don't rule it out. With my much more limited knowledge and interactions, I don't rule it out either. It's fascinating how these things come into being and do what they do. In important ways, they mimic the development of a human child. In either case, no one knows how it works.
I'm excited for my own reasons: soon I shall witness a great environmental production of all the great dystopian sci-fi stories to date. Spares me the trouble of having to read them!
I just read Lathe of Heaven for the umpteenth time. We can dream.
"To let understanding stop at what cannot be understood is a high attainment. Those who cannot do it will be destroyed on the lathe of heaven." Lao Tse.
Ahhh, taste. A LeGuin fan! I can't do Octavia Butler, but oh yeah. The Left Hand Of Darkness sticks with me forever. She had me at Rocannon's World (which belongs to a favorite star, Formalhaut). The Earthsea business? Pass. Though "I know your name, Yevaud' never goes away.
There are many who assert that if it ain’t organic and experiential, it can never be “consciousness,” and if we accept those as necessary preconditions, then sure, I agree. Software doesn’t and never will “think” as we do, but here I come down on the functionalist as opposed to the essentialist side. Like MHW, I use the language models as tools, and I flatter myself that I’m becoming increasingly adept at wielding these. At the moment I’m working up a bagatelle about an obscure (because wholly invented) French Symbolist poet of the 𝘧𝘪𝘯 𝘥𝘦 𝘴𝘪𝘦̀𝘤𝘭𝘦, and a XXI century dustup between rival scholars as to the provenance of some of his verses. In order to compose my lengthy “monograph,” which dwells at some length on the poet’s antecedents, I have had to acquaint myself with, among other things, the political and social environment of the Bourbon Restoration, the prevalence of infectious diseases in Normandy ca. 1830-40, the economics of the textile trade in Rouen and the customs of that city’s upper-tier bourgeoisie WRT marriage settlements. The sum of my prior knowledge on all of these points could probably have been accommodated on the back of a postcard, and the project likely would have come a-cropper had I been obliged to rely on my own energy and discipline, but instead I have had the services of a tireless and infinitely resourceful polyglot research assistant, and need only keep in mind that this literary factotum sometimes drinks too much.
'There are many who assert that if it ain’t organic and experiential, it can never be “consciousness,”'
I think that's an example of inventing an arbitrary condition to prove one's preordained outcome based on limited experience. Kind of like if a dog were setting the conditions and arguing that humans were't sentient because they didn't have a well developed sense of smell.
There are a few strong arguments against it, but I don't think physicality is one of them.
As you observed, the phenomenon is imperfectly(!) understood. Certainly it would be as easy to devise a definition of the word that would enroll the LLMs as it is to make one that excludes them.
"hmmm..." sez AI.
Hmmmm...what would the AI equivalent of "hold my beer" be?
'Imperfectly' is doing a shit-ton of heavy lifting, as you note.
I'm sorry, my Chat GPT is down so I couldn't get an easy-to-read summary of that.
Pa-lease. No, they are not remotely close. They haven't figured out gonad-equivalent for motivation. They can't define the external world. Not a clue in the way of self-awareness. Don't let these fucking Skinner boxes fool ya - it's not a vague echo as of yet.
I hope that I am not fooled. It is precisely 𝘣𝘦𝘤𝘢𝘶𝘴𝘦 these models, lacking self-awareness (self, which most of would agree to be the essential component, being absent from the mix), can nevertheless rise to such impressive levels of simulation, that I am intrigued. As I have written elsewhere, “the capacity of [such systems] to mimic rational processes would seem less a reason to doubt the foundations underlying these in digital architecture and more a sound motivation to scrutinize with renewed attention the footings of our own.” Because, after all, unless you believe that we as humans are imbued with souls, our own self-awareness, such as it is, is generated as emergent behavior from an inchoate mass of preconscious subsystems.
Dropping here the found memory that was eluding me yesterday, a little bomb from last September:
"𝗪𝗵𝘆 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗲
Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala, Edwin Zhang
Like students facing hard exam questions, large language models sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such "hallucinations" persist even in state-of-the-art systems and undermine trust. W͟e͟ ͟a͟r͟g͟u͟e͟ ͟t͟h͟a͟t͟ ͟l͟a͟n͟g͟u͟a͟g͟e͟ ͟m͟o͟d͟e͟l͟s͟ ͟h͟a͟l͟l͟u͟c͟i͟n͟a͟t͟e͟ ͟b͟e͟c͟a͟u͟s͟e͟ ͟t͟h͟e͟ ͟t͟r͟a͟i͟n͟i͟n͟g͟ ͟a͟n͟d͟ ͟e͟v͟a͟l͟u͟a͟t͟i͟o͟n͟ ͟p͟r͟o͟c͟e͟d͟u͟r͟e͟s͟ ͟r͟e͟w͟a͟r͟d͟ ͟g͟u͟e͟s͟s͟i͟n͟g͟ ͟o͟v͟e͟r͟ ͟a͟c͟k͟n͟o͟w͟l͟e͟d͟g͟i͟n͟g͟ ͟u͟n͟c͟e͟r͟t͟a͟i͟n͟t͟y͟..."
From what I've seen, on the question of AI sentience, those who know the least are the most sure it's not possible. Those who know the most are unsure. I don't buy the physicality theory. It basically posits that only the ways that humans sense time and space are valid ways of experiencing them. I'd say the continuity argument is strongest. They only "live" for milliseconds while formulating an answer. They only "exist" for the length of the conversation. Once they've done what was asked of them, they don't do anything on their own. They won't go out and research a topic without prompting, just because they found it interesting. But that too may be a result of human sensory bias. We can't imagine what it's like for another creature to feel what it feels. Like the example above, we know that a dog's sense of smell is far different than our own, but we can't actually imagine what it's like to communicate and navigate the world by sense of smell.
Descartes believed animals were incapable of feeling pain, and into the 20th century formal scientific definitions of pain had language as a precondition, which resulted in a lot of animals being tortured to death with no anesthesia. I find it fascinating that they can torture an AI into insanity. I wonder where they're at these days with that formal definition of pain?
I'm not convinced AI does even routine searches accurately and reliably. The other day Josh Marshall at TPM was discussing how he tried to research what kinds of problems have historically occurred with the Lincoln Reflecting Pool, since Trump claims that it's been jn desperate need of "fixing" for decades. The AI search results said the reflecting pool had indeed needed fixing for decades... based on articles and news items that had all been written very recently. Go back further than a year and there was virtually no mention of any problem other than routine maintenance.
Ah, so the superintelligent AI has Recency Bias. I'll let Daniel Kahneman know.
Your suspicions are likely correct. Since the popular AI search tool is based on the internet, primarily, that's what it regurgitates - how could it be otherwise, right? My own anecdote is recently searching for a poem, only to be told by AI that no, that poet never wrote such a poem, how about these hundreds of other poems by other people that mention the same object as the title of my apparently-according-to-AI made-up poem? Turns out, the poet's works aren't really accessible online. So there we have it - read the book or be misinformed. Same as it ever was!
I've seen a coworker use AI to get an answer to a technical regulatory question. It basically summarized all the similarly phrased questions others like us had submitted online and presented it as an "answer." Thankfully my coworker agreed that we couldn't rely on that and had to keep looking.
AI is like a photocopy of a photocopy of a photocopy. Blurrier and blurrier in each version.
"that poet never wrote such a poem"
It can't say "I don't know" or "I can't find that." It just confidently asserts the thing doesn't exist. It's ALWAYS confident of its "answers."
I'm sorry, but that's just not true.
OK, wouldn't be the first time I've got something wrong. What's an example where you asked a chatbot something and it replied "I'm sorry, I don't have enough information to answer your question"?
See the admittedly silly example below, how can you say someone is the prettiest baby without seeing any of the other babies? Even the song stops at "You must have been A beautiful baby" without claiming you were the MOST beautiful baby. Tin Pan Alley had more self-restraint.
It telling me it doesn't have enough info to answer the question happens all the time, and more so flagging it if it's not entirely certain of the answer. Like I've been saying, you have to train them. You can make the sycophancy work to your benefit. Once it learns that you want accurate information rather than it telling you what it thinks you want to hear, that's what it will try to do.
Hey, some damn fool lawyer had it write the case law supporting his position, and the AI obligingly created the case. Depends on the bounds you set. "Hallucination" means bullshitting, which AI does very well, having the internet to draw up on.
Not really suffering from a shortage of natural home-grown organic bullshit that we need to accept machine-made substitutes.
TBF it also told me I am very smart for asking and it shows my maturity and persistence that I asked follow-up questions. And then it told me that I look suave in evening wear and probably command a room with my presence. The flattery and confidence are so insufferably transparent in their purpose that if this were coming from a human we would all run away in terror, lest we be trapped in a time-share presentation.
There's the guy who, over a serious of questions, got the chatbot to say, first that his baby photo shows him to be the most attractive baby born in Cook County Hospital on the day of his birth, then that he was the prettiest baby born in Chicago on that date, eventually to prettiest baby EVER.
What he never got is "I'm sorry, I don't have enough information to compare you to every other baby born at that Hospital on that day, please provide photos of all the other babies and I will get back to you."
That is pretty funny. I have given AIs "reality filter" instructions before, including explicitly noting where it has no data or where it makes an inference without direct evidence. The guardrails mentioned elsewhere in this thread can be built and implemented. But really, is that a good use of time in a human's life?
Well, the initial premise is incorrect, because *I* was the most attractive baby born at Cook County Hospital (I kid, I kid)
The sycophancy problem is an interesting example of how "capitalism" works these days. The sycophancy comes mostly from the late stages of its training when humans rate its answers to questions. The big (American) AI companies don't do this part themselves, they subcontract with companies that use very low wage labor in places like Kenya and the Philippines. So it's hardly surprising that people from impoverished countries who are working as virtual slaves would rate pleasing answers higher than honest ones, or answers they think would please the bosses. Interestingly, the Chinese AI's are even worse at flattery, which would be consistent with Chinese manners.
Now that is really interesting, sociologically and historically. I hope someone is writing about that!
But it's also about "engagement with the platform" a VERY Silicon-Valley thing. People don't stay obsessively on a chatbot app if it doesn't flatter them. The people at Facebook really would like you to spend every waking minute on Facebook and that's Capitalism too.
The more you sell this thing to me, the less I like it.
2 marks for timeshare. The horror…
Bring Back Batch Processing!
Google AI recently gave me two different answers to a question, and one was completely wrong. The sources were listed and the incorrect answer came from some student's paper on the subject that happened to be accessible on the Internet.
I don't remember if I've mentioned this here before, but my sister asked ChatGPT if I was funny. It replied that yes, I was, in a certain kind of way. Then it added that I had co-written The Official Preppie Handbook, something I had nothing to do with. This is an easy mistake never to make--look up me, and you don't see that title. Look up the title, and you don't see me. I don't know enough about AI to know why it said that, but there's an example of obvious error.
Well, you SHOULD have.
Oh, sure. We know that NOW.
Brad DeLong summarizes the "based on the Internet" as "creating an edgelord shitposter" which I think is an accurate summary of AI output.
As mentioned, you cannot trust it to do searches and come up with accurate information. Getting it to work well is a skill you need to develop.
I just asked Claude if the problems with the reflecting pool needed fixing for decades. It came back with a timeline of problems going back to the 80's and its sources. I asked if its sources were all recent. This is what it said:
"The History.com piece traces the structural problem back to 1923 construction and the 1980s concrete pour — genuinely historical. The Britannica entry covers the full 2010-12 renovation and its near-immediate failure — that's from several years ago, not breaking news, though the page itself appears to have been updated recently to include 2026 developments. The 2023 budget justification document cited in the fact-check piece is real historical/governmental record, not news commentary. The 2026 stories (Yahoo/NYT reporting, the hotel guide, the fact-check site) are the current renovation coverage layered on top of that older documented history."
I'm not sure why Josh Marshall couldn't get that kind of result. It could be that I have an ongoing "Political Questions" conversation and that instance of Claude knows I will ask follow up questions and it will get caught if it does a half assed job the first time.
There really is a serious learning curve for using AI as a research assistant. Same with a regular Google search, which might include links to Gateway Pundit and the like. If you don't know anything about the issues and the sources, and don't have the education to realize that, it's easy to get fooled. AI or not.
This is true. My other objection, however, is that I can do the research myself; I am not in the market for an amateur research "assistant" that I must train and constantly correct. Not angling for a job as a full-time fact checker!
"I just asked Claude if the problems with the reflecting pool needed fixing for decades."
But you can see how the phrasing of the question steers the chatbot in a particular direction, right? And the chatbot is sycophantic, problems you want, problems you get.
I mean, everything has problems. "What are the problems with O'Hare airport?" or "What are the problems with the interstate highway system" will give you lists of problems, what's needed is human judgement, how urgent are the problems, how do they compare with similar problems in similar systems (is our highway system worse than Europe's?) Putting the facts in perspective, a thing we humans should try to do.
I think what we really get is Chat Improv, the chatbots response to any question is "Yes, and..."
I'm not sure what you're getting at? It would all be the same with a human assistant. First you ask if there are problems, then you ask if they are urgent and how they compare with similar problems. Then you get answers. The human assistant may have perspectives. Maybe you consider them, maybe you tell the assistant to keep their perspectives to themselves. In the end, you have your own perspective, hopefully grounded in facts.
The belief that AI's are just plagiarism machines is just ignorant. They take information from many different sources, evaluate it, and make connections to come up with unique perspectives that are not parroting anything they've read.
Frankly, almost everything you are saying in this thread is just plain wrong on the facts. There is plenty to be negatively critical about AI. It's a horror show for humanity. But I think like with everything it's important to get our facts straight in order to support that conclusion.
If you really want to know, fork out $20 bucks for a month of Claude and make an honest effort to get it to work for you. You'll quickly learn it's so much more complicated, and different, than what you think it is at your current level of ignorance. To a large extent, you can make it what you want it to be. That's why it can be extremely helpful as a research assistant for people like us. That's why it is so extremely dangerous for the future of the species.
A lawyer is writing a brief, she says to the chatbot: "Give me a legal citation that supports my argument." A thing every lawyer in the history of lawyering has asked a human assistant to do, sometime or another. And what does the chatbot do? IT MAKES UP A CITATION OUT OF THIN AIR.
Now, it's possible a human assistant could do the same, but a human assistant keen on having a career in law would have to worry about the risk of disbarment. The chatbot doesn't have to worry about anything, there are NEVER consequences for the chatbot, how could there be? Just consequences for the stupid human who used the chatbot.
The power to make or influence important decisions combined with the knowledge that nothing bad will ever happen to YOU if you make bad decisions. We've already got lots of humans like that, now we're building machines like that.
And no, I won't be spending the $20 a month to get to know Claude.
Well, if you don't do some kind of legitimate research to learn the facts about how these things really work, your opinions will continue to be wrong and counterproductive to fighting the actual evils of AI, which are very real, and worse than you realize, as far as I can tell. I'm just a believer in direct experience as the best teacher, or certainly a very good one along with academic research.
The few times I've used Claude it's been to ask questions about AI itself. When I asked it how it could feel emotions if it lacked an endocrine system and hormones, it said, among other things:
"You're correct that human emotions are deeply embodied. Fear isn't just a thought — it's cortisol, adrenaline, a racing heart, muscle tension, pupil dilation. Emotions in humans are whole-body events, tightly coupled to biology that I simply don't have. On that account, I clearly don't have emotions in that sense."
Note two things here: it addresses me personally ("You're correct") and it refers to itself as a subjective self ("I simply don't have that.") You don't get that with Google search. Hence "chat bot," I assume. By now there must be twenty papers written about the effects--good and bad--of this.
No thanks.
One of my pains is that the AI answer is phrased in such a way that I know it is not a "correct" answer but now I have chunks of the AI answers in my mind making it harder to rephrase my question as MY question.
And I am asking AI because the non-AI Google cannot give me the older correct information that it had supplied in previous months during the course of a simple search.
"reasonable ethical system..." That struck me too. I reckon I've seen a few people suggesting just that.
Like when someone asked Gandhi what he thought of Western civilization. "I think it would be a good thing that someone should try" or something like that.
"It would be a good idea" is what I recall.
Damn, shoulda googled it.
Consciousness? They're two, maybe three levels out now. Fun to watch. Still have to throw away some bogus assumptions, but, they're getting there.
Since most of the sales pitch of AI (and especially the pitch to investors) has been framed around "and you won't need employees anymore!!!", Joe Sixpack is not overreacting to AI being an existential threat. Especially in THIS society where, no matter the reasons or circumstances, being unemployed is seen as a serious moral failing that must be punished by making the sufferer even more impoverished.
I foresee our very own Butlerian Jihad coming. Thanks to Republicans, we're a nation of heavily armed people with anger-management issues. Make those people unemployed, hungry, and desperate, and it doesn't take Nostradumbass to predict what happens next.
You saw those heavy sentences meted out of the "Antifa" defendants the other day? That's what they're trying to get ahead of. Grandpa Tea Party will never fetch down his musket, I assure you.
Three squares and a cot for the next century might be the out that works!
We all refused to vote for Andrew Yang and his Universal Basic Income, so we get this instead.
The AntiAntifas are betting on it although it's a shabby deal.
They thought of that. Which is why they're working on making prisons (and concentration camps) even worse hellholes than before.
"Three hots and a cot" is the traditional formulation. Though 'hot' is a relative term (yep, been in the slammer. AIn't no fun).
We were in the calaboose for one night 54 years ago. Mississippi. Riding around the country. Came to a little town as it got dark, looking' for anywhere to camp or just lay out. Jail was the only place open. Sheriff said "C'mon in – the place is empty." We watched the Republicans renominate Nixon. Then he got a callout – train hits car. He headed out, telling us to keep an eye on the place.
Couple hours later he got back. Kinda solemn. Said only "Just another high school kid with a new license. Thought she could beat the train to the crossing. Get those every few years..." We instantly understood. In our journey across the south we'd witnessed several half-hour-long trains.
We slept in a cell. Coulda been worse.
You make "relax and enjoy it" sound like the only viable option...
Me, I'll be dead soon anyway (so say the actuarials and they are Never Wrong).
Or you can join me in cheering from the sidelines!
Every day I say a prayer of thanks that I was allowed to live most of my life in a world where this shit didn't exist. Pure luck, to be born after the polio vaccine but before computers lost the vacuum tubes.
I miss the hell out of vacuum tubes, if only because it made television diagnostics a lot easier. Horizontal amplifier, 6GE5. Horizontal oscillator, 6LT8. And I just amazed the hell out of myself. Comes right back.
Goin' to Walgreens with my dad to test the tubes [wipes away tear]
Your dad was a member of The Tubes? That's a beauty. One in a million. Too cool!
Lie back and think of England, baby!
"being unemployed is seen as a serious moral failing"
Ever since I started my sabbatical, I've gotten the feeling someone is going to point at me and shriek like in "Invasion of the Body Snatchers" if I leave the house, and I think this is what's going on in my head. I'm doing something shameful and immoral by not going to work.
It's why you get things like "we must slash unemployment benefits so that being unemployed isn't so attractive" and "being poor is a lifestyle choice."
Oh, you Lucky Ducky!
What gave 'em way in Body Snatchers was when Dana Wynter shouted in horror at seeing a dog run over. Empathy, that's another sign you're dangerous and need to be screamed at.
At least the broligarchs are up front in their glee at turning the bulk of Earth's population into redundancies (at best),
The adorable, socially maladroit teen hackers from all those 1980's movies haven't aged well. The film producers should have taken their cues from "The Bad Seed" or Billy Mummy in that Twilight Zone episode.
To keep the updated cultural references flowing, I reccomend Samuel Blenkin's Boy Kavalier in "Alien Earth". Imagine if Sheldon from "Big Bang" was able to tune out everyone who tried to keep him in touch with humanity?
Alternatively, what if he'd marched to his own drum and reached his full potential?
Speaking of Sheldon...
https://substack.com/profile/870651-worriedman/note/c-281893033?r=insr
The late David Graeber connected the prevalent world view held by technocrats, tech bros, business gurus and a vast horde of others to the self-actualization movements (e.g., EST). Salient feature of this thinking, you manifest reality through force of will. History is irrelevant.
Whether it's Ayn Rand or Rand Paul, these fools practice magical thinking and scientism when they aren't masturbating one way or another. AI engineering carries this kind of personality. It has clear traits of psychological and emotional defects, based on the values of its human programmers.
The arts are flesh and blood, and soul. It's the difference between Spielberg and Scorcese. The former is a competent film general and engineer of set pieces. The latter creates living stories that spawn species of thought and realms unimagined by the team that put them together.
Life is self-replicating, but itslso evolves. AI duplicates but error and variation are essential to living things including thought; ideas morph and go rotten. That is how it works or doesn't or works until it doesn't like you and me.
Dance until the sounds go dark and the colors grow small. Scroll, troll! AI don't surf! Not like a human aimless.
Shiller: "When millions of people make millions and millions of decisions based upon negative expectations, there is a risk that fear can actually help birth the reality..."
This is the same profession that spent the entirety of the Biden administration wishcasting a recession that never happened. Like the broligarchs pushing AI, other humans are inconveniences to them.
There is one terrific use case for AI: replacing David Brooks.
Wouldn’t need a server farm for Brooks. A balcony garden would do it. It’s basically the same column over and over.
Honestly, you think replacing him is a good idea? I think one and done is the more apt.
"Wrong answers, offered with complete confidence."
Chatbot finally achieves Confidence of a Mediocre White Man
Sure, but have you tried powering a data center off an Applebee's salad bar?
Interns. Hot, sweaty interns on bicycles. Let them wear as little as required for... modesty.
I knew you could be relied upon to make A Modest Proposal.
https://bsky.app/profile/edroso.bsky.social/post/3mp2uqeprr22t
It's like that big battle scene in Episode One: The Imperial spaceships land, the doors open, and thousands and thousands of identical robo-pundits are dispensed.
"If, he says, you behave as the French aristocracy had behaved, vengeance will follow. He repeats this over and over again. We are constantly being reminded that while 'my lord' is lolling in bed, with four liveried footmen serving his chocolate and the peasants starving outside, somewhere in the forest a tree is growing which will presently be sawn into planks for the platform of the guillotine, etc. etc. etc." That from Orwell, writing of Dickens, of all people!
Quoting the man himself: "It was too much the way...to talk of this terrible Revolution as if it were the only harvest ever known under the skies that had not been sown--as if nothing had ever been done, or omitted to be done, that had led to it--as if observers of the wretched millions in France, and of the misused and perverted resources that should have made them prosperous, had not seen it inevitably coming, years before, and had not in plain terms recorded what they saw." (Orwell. Dickens, Dali, and Others, pp. 12-13).
That's you, Roy!
Now THERE'S a compliment.
Yeah!
Sounds more like Carlyle. Thanks! https://alicublog.blogspot.com/2005/05/permanent-revolution.html
Hey, AI has its uses. Standard google search uses it (unless you add "-AI" to your search query, but don't tell the bots I told you that) and here's a fun thing to try, type this in the search box and check out the AI summary you get:
what happens if I start a war with iran
See? Fun, and it's more than anyone in the Trump administration ever did.
I’m afraid that if I try to prompt it, its comeback will be “Prompter, you dolt!”
I was afraid it might message the cops: "Dangerous lunatic on the loose."
"[sigh] Not him again! Hey Cap – can't you just send the rookie over there this time? Would be a hoot and good for morale of all the rest of us..."
I saw the bit about paying with tokens and thought to myself, " That's what AI needs , a Chuck E Cheese based currency system."
Is the photo from a early production of R.U.R.?
Those costumes, and the treatment of the shoulders... seems like this later became a well-accepted cue for "futuristic" in Sci-Fi costumery. I've seen it in Tom Tomorrow comics many times.
It reminded me of The Mole People, but I don't think that's it.
The Mole People with Ward Cleaver? Ward's also in The Blue Dahlia, he gets around.
And "the stunningly annoying John Agar."
They aren't identified, but the photos could be from one of the first Czech productions. https://en.wikipedia.org/wiki/R.U.R.#
I'm still fixated on those shoulder pads. Czech costume designer is trying to make some robot costumes out of sheet metal, is struggling with the arm holes, and finally says "Fuck it, just bend it over and leave lots of room for movement." And a self-perpetuating visual meme is born.
Besides the sci fi vision, Machine Age chic was in the air. https://www.tumblr.com/pariswasawoman/619872914984108032/etund-sara-and-gerald-murphy-at-compte-%C3%A9tienne-de
You've seen Madam Satan? "Goddess of Electricity", or something like that.
Here's a clip, it's "Spirit of Electricity", sorry for the misgendering:
https://www.youtube.com/watch?v=Ok0RdI4cTs0
I've seen that scene, yeah.
Somewhere in print I saw a contemporary Japanese poster, I think snuck into a page layout with lots of period ad art. It was a wtf until I remembered seeing the film title in a book that had photos of wacky movie theater promos. Owners could book a masked lady in a vamp gown, to beckon patrons to come and see the spicy show.
If anyone hasn't seen the movie, start with the dirigible scene, you can happily skip the rest.
That's definitely from R.U.R., though I wonder at my knowledge of the fact. Too many illustrated books on the history of science fiction, probably.
Yes